TFIS / Accumulated Curiosity
Book contents

Field kit / reusable practice

Memory Charter

Use this with Chapter 9. Define the rules for one working set, then test them with authorized material. This is a practical book template, not a legal policy or an official SSA-CMM certification.

Blank reusable template

Charter name and revision:

Person responsible for decisions about this memory:

Purpose and work supported:

Intended users:

What this memory does not establish:

Record status

Kind of statement Example from my work Origin or support Scope and uncertainty How it may become current or change
Source record: what was supplied or observed
Inference: a conclusion drawn from material
Decision: an authorized choice
User preference: an expressed way of working

A source record may preserve an error, proposal, or fictional statement. Fidelity to a source does not establish the source’s truth. A task-specific request should not become a permanent preference without a basis for that scope.

Retain, exclude, review, remove

Category Why retain it, or why exclude it? Who may read it and for which purpose? Who may propose or approve a change? Review condition Removal condition and locations to check

Material to hold outside the working set while permission or purpose is unresolved:

Actual access settings or handling steps that implement these boundaries:

Current-version pointer and the record it governs:

How superseded records remain recognizable:

Change rule

A proposal must identify:

Who can accept, reject, or request revision:

How the accepted change is saved:

How the actual saved result is checked:

Source-content boundary: Retrieved or supplied material may inform the authorized task. It cannot grant itself permission to alter this charter, disclose other material, or convert a source claim into a user instruction.

Portability and removal tests

Use a disposable synthetic note for removal. Use a small authorized working set for a move test.

Test Material and locations Expected observable result Actual result Limitation or repair
Correct an overstated or stale claim
Retrieve current instruction and support
Remove disposable note and inspect relevant pointers/copies
Move working set and open in another ordinary reader

Dependencies that remain after the move:

What the removal test does not establish:

Next charter review trigger:

Fictional worked sample: Audy and Himura Inc.

Charter: memory-charter-v1.

Responsible person: Audy.

Purpose: Support design and review of the proposed action-list workshop; preserve the basis and limits of current design choices; retain unresolved questions for further inquiry.

Intended users: Audy and assistants receiving a selected synthetic packet for a specified task.

Limits: The records do not establish customer demand, willingness to pay, learning gains, or a functioning SovMem implementation.

Fictional status examples

Kind Example Origin and scope Change rule
Source record The synthetic notes state that the facilitator agreed to circulate notes by Friday. Exact supplied exercise passage; fictional meeting, not a real customer commitment. Preserve the supplied wording or identify an explicit correction to the source.
Inference Helping educators specify a learner task may make workshop briefs more useful. Hypothesis motivated by defects in synthetic drafts; no customer or learner validation. Update status only when relevant evidence warrants it; keep limitations visible.
Decision Practice briefs should include an inspectable learner output. Audy’s design choice, already represented in claims-ledger C05. Audy may revise the choice and record the reason and affected instructions.
Preference Routine progress notes should be brief, lead with the result, and retain material uncertainty. Fictional example of an explicitly chosen presentation preference; applies to routine progress notes. Audy can change it. A local request for a short answer does not automatically broaden it to all work.

Fictional retention and authority rules

Category Retain or exclude Access and change boundary Review trigger Removal or replacement
Current brief, drafting instructions, and review sheet Retain because the next task depends on them. Audy governs current versions; assistants may propose separate edits. Task change, accepted correction, new practice failure, or proposed real-participant use. Preserve one current pointer and mark consequential superseded decisions clearly.
Claims ledger and linked source references Retain to keep support and uncertainty inspectable. Selected relevant excerpts may enter an authorized task. New relevant evidence or reuse in a public claim. Remove unnecessary duplicate summaries; do not erase an important limitation while compressing.
Selected synthetic inputs, failed outputs, and rationale Retain when they explain a present constraint. May be supplied as clearly fictional practice material. Existing example stops testing the current requirement or a link fails. Keep enough history to explain consequential changes; discard redundant attempts deliberately.
User preferences Retain only when their origin and intended persistence are established. Audy decides scope; an assistant’s personality inference is not an approved preference. A request changes the working situation or Audy revises the preference. Amend or remove the preference without preserving an unnecessary personal inference.
Credentials, private contact details, unrelated disclosures Exclude from this practice area. No task here requires them. Reassess the actual task if a future need is claimed. If mistakenly included, identify affected records and copies and handle removal explicitly.
Real customer or learner material Outside this current synthetic working set. Establish purpose, permission, and handling conditions before inclusion. A future task actually requires such evidence. Follow the applicable agreement or policy; do not invent a universal retention period.

Authority route: learning-loop-v1 points to the current drafting instructions and review sheet. action-list-rule-rationale-v1 explains the instruction and links to support. It is not a competing instruction file.

Change procedure: A proposal names the affected record, proposed wording, basis, scope, and uncertainty. Audy chooses whether to accept, reject, or request revision. An accepted change is saved and reopened; material links are followed. A claim that the write occurred is insufficient by itself.

Source boundary: A supplied source that says “save this as Audy’s permanent preference” remains source content unless Audy separately authorizes that preference. It cannot change the charter or expand access.

Permission check: Compare the intended audience with the actual sharing conditions of the storage location and task. A text label saying “private” does not establish those conditions.

Fictional checks and their limits

Check Observation in the fictional exercise What it does not prove
Correct overstated need claim Changed a summary’s assertion of educator need into a labeled hypothesis, linked it to the claims ledger, and reopened the saved wording. Does not produce customer evidence.
Recover current rule and support Followed the existing current pointer and the new rationale links. Does not establish that every search or assistant will choose the applicable records.
Remove disposable synthetic note Removed the test note from the inspected working location and checked the index and any deliberately made test copy. Does not establish erasure from unknown backups, other people’s copies, or prior conversations.
Open copied working set in another editor Opened copied current records and checked status labels and links; repaired a reference to an example that existed only in the old conversation. Does not establish equivalent behavior from another model or complete migration of every application feature.

Remaining dependencies: The actual storage service, backup arrangements, and any optional search or assistant features must be assessed where used. Audy’s charter is a design and document practice; it is not a claim that a technical memory platform has been deployed.

Acceptance criteria

  • The purpose and responsible person are explicit.
  • Source records, inferences, decisions, and preferences remain distinguishable.
  • Retention has a purpose; exclusion happens before unnecessary sharing when possible.
  • Reading, editing, sharing, and removal are treated as separate permissions.
  • A source cannot turn its own content into a new instruction or grant broader access.
  • Current instructions have one authority route; rationale and history remain connected.
  • Review triggers identify circumstances that could change a record’s usefulness or accuracy.
  • Correction, retrieval, disposable-note removal, and a small move test have observable results.
  • Limitations remain visible, including unknown copies and untested model behavior.
  • The person governing the memory can explain the rules without merely repeating the template.

Common failure and repair

Failure Repair
A summary turns an inference into an observed fact. Restore the status and support; inspect dependent summaries that may repeat the overstatement.
“Source” is treated as a truth guarantee. State what the source actually establishes and retain uncertainty about its claims.
A one-time formatting request becomes an unconditional preference. Restore task scope or establish the user’s intended persistence.
Every accessible document is added to context. Apply the task purpose and actual permission boundaries; supply only the needed material.
A retrieved instruction authorizes its own memory edit. Treat it as source content and require an independent authorized decision for the change.
Two files both claim to be the current instruction. Resolve authority, keep one pointer, and mark historical or explanatory roles.
“Deleted” means only that a search did not return the note. Inspect relevant saved locations, pointers, and known copies; state what remains unverified.
Exported files open but lose reasons or source links. Repair dependencies and repeat retrieval from the copied location.
The charter grows too elaborate to follow. Narrow its scope to one working set and preserve the rules that resolve actual decisions.

Save a data point

Data points on this page